@empiricalrun/test-gen 0.46.7 → 0.46.9

package/dist/agent/master/run.js

@@ -8,25 +8,25 @@ const utils_1 = require("../../actions/utils");
 const logger_1 = require("../../bin/logger");
 const constants_1 = require("../../constants");
 const errors_1 = require("../../errors");
-const human_in_the_loop_1 = require("../../human-in-the-loop");
 const page_1 = require("../../page");
 const reporter_1 = require("../../reporter");
 const session_1 = require("../../session");
-const env_1 = require("../../utils/env");
-const browsing_1 = require("../browsing");
 const utils_2 = require("../browsing/utils");
 const skills_retriever_1 = require("../codegen/skills-retriever");
 const run_1 = require("../planner/run");
 const run_time_planner_1 = require("../planner/run-time-planner");
 const utils_3 = require("../utils");
 const action_tool_calls_1 = require("./action-tool-calls");
-const element_annotation_1 = require("./element-annotation");
+const execute_browser_action_1 = require("./execute-browser-action");
+const execute_skill_action_1 = require("./execute-skill-action");
 const next_action_1 = require("./next-action");
-const with_hints_1 = require("./with-hints");
 const MAX_ERROR_COUNT = 2;
 function getPageVariables(stateVariables) {
     const keys = Object.keys(stateVariables);
-    const pageVariables = keys.filter((key) => key.endsWith("Page") || key.endsWith("page"));
+    // This checks for whether page.url() exists, which is true for all pages
+    // created by playwright actions.
+    const pageVariables = keys.filter((key) => typeof stateVariables[key] === "object" &&
+        typeof stateVariables[key]?.url === "function");
     const pages = pageVariables.reduce((acc, key) => {
         acc[key] = stateVariables[key];
         return acc;
@@ -110,13 +110,12 @@ async function createTestUsingMasterAgent({ task, page, testCase, specPath, opti
                 failedActions,
             },
         });
-        // Provides a plan to master agent to execute, also verifies whether task is done or not
         const plannerResp = await (0, run_time_planner_1.runtimePlanner)({
             trace: masterAgentSpan,
             task,
             successfulActions: [...masterAgentActions],
             pages: getPageVariables(actions.getStateVariables()),
-            currentPage: (0, utils_1.getPageVarName)(),
+            currentPage: testGenPage.name,
         });
         isGivenTaskDone = plannerResp.isDone;
         if (isGivenTaskDone) {
@@ -131,9 +130,7 @@ async function createTestUsingMasterAgent({ task, page, testCase, specPath, opti
         }
         // inject scripts in the page
         await (0, utils_2.injectPwLocatorGenerator)(testGenPage.pwPageInstance);
-        const buffer = await page.screenshot({
-        // path: `screenshots/screenshot-${screenshotIndex++}.png`, // enable this and screenshotIndex var for local debugging
-        });
+        const buffer = await page.screenshot();
         const testGenReporter = new reporter_1.TestGenUpdatesReporter();
         await testGenReporter.sendCurrentView(buffer);
         const pageScreenshot = buffer.toString("base64");
@@ -143,31 +140,29 @@ async function createTestUsingMasterAgent({ task, page, testCase, specPath, opti
             break;
         }
         // Provides next action that needs to be taken
-        const toolCall = await (0, next_action_1.getNextAction)({
+        const nextAction = await (0, next_action_1.getNextAction)({
+            page,
+            pageScreenshot: [pageScreenshot],
             task,
             executedActions: masterAgentActions,
             failedActions,
-            pageUrl: page.url(),
             trace: masterAgentSpan,
             llm,
             options,
-            pageScreenshot: [pageScreenshot],
             actions,
-            actionTypes: action_tool_calls_1.ActionType,
             disableSkills,
-            page,
             logger,
         });
         if (await (0, session_1.shouldStopSession)()) {
             break;
         }
-        if (toolCall) {
-            if (toolCall.actionType === "unknown") {
+        if (nextAction) {
+            if (nextAction.actionType === action_tool_calls_1.ActionType.UNKNOWN) {
                 logger.error("Agent is not able to figure out next action since element is not visible on screen.");
                 await testgenUpdatesReporter.sendMessage("Agent is not able to figure out next action since element is not visible on screen.");
                 break;
             }
-            const args = (0, utils_3.parseJson)(toolCall.toolCallArgs);
+            const args = (0, utils_3.parseJson)(nextAction.toolCallArgs);
             const masterAgentActionSpan = masterAgentSpan?.span({
                 name: "master-agent-execute-action",
             });
@@ -178,124 +173,35 @@ async function createTestUsingMasterAgent({ task, page, testCase, specPath, opti
             try {
                 void testGenReporter.sendMessage(output.reason);
                 logger.log(`Next Action: ${output.action}`);
-                if (toolCall.actionType === skill_1.SKILL_USAGE) {
-                    if (!(0, env_1.isRunningOnCloud)()) {
-                        const { skillDetails } = (0, skill_1.extractSkillFromArgs)(args);
-                        const feedback = await human_in_the_loop_1.humanLoop.getFeedback({
-                            message: `Should I use this skill?\n\n${skillDetails?.usageExample}\n\nfrom: ${skillDetails?.filePath}\n\nAnswer with "yes" or "no"`,
+                switch (nextAction.actionType) {
+                    case skill_1.SKILL_USAGE: {
+                        const code = await (0, execute_skill_action_1.executeSkillAction)({
+                            actions,
+                            nextAction,
+                            trace: masterAgentActionSpan,
                         });
-                        if (feedback.toLowerCase().trim().includes("no")) {
-                            throw new errors_1.HumanApprovalDenied("Skill usage approval not provided");
+                        if (code) {
+                            generatedCodeSteps.push(code);
                         }
+                        break;
                     }
-                    const code = await actions.executeAction(toolCall.actionType, args, masterAgentActionSpan);
-                    if (code) {
-                        generatedCodeSteps.push(code);
-                    }
-                }
-                else {
-                    let shouldTriggerHintsFlow;
-                    let hintsExecutionCompletion;
-                    let elementAnnotation;
-                    const actionType = toolCall.actionType;
-                    let preference = {
-                        actionType: "all",
-                    };
-                    if (useActionSpecificAnnotations && (0, action_tool_calls_1.isValidActionType)(actionType)) {
-                        switch (actionType) {
-                            case action_tool_calls_1.ActionType.FILL:
-                                preference = {
-                                    actionType: action_tool_calls_1.ActionType.FILL,
-                                };
-                                break;
-                            case action_tool_calls_1.ActionType.ASSERT_TEXT:
-                                preference = {
-                                    actionType: action_tool_calls_1.ActionType.ASSERT_TEXT,
-                                    assertionText: args.assertion_text,
-                                };
-                                break;
-                            default:
-                                preference = {
-                                    actionType: "all",
-                                };
-                        }
+                    case action_tool_calls_1.ActionType.OBSERVATION: {
+                        output.action = args.observation;
+                        logger.log("Observation: ", output.action);
+                        break;
                     }
-                    let { annotationKeys, annotatedPageScreenshot, annotationBuffer } = await (0, element_annotation_1.getAnnotationKeys)({
-                        page,
-                        preference,
-                        options,
-                        trace: masterAgentActionSpan,
-                    });
-                    if (annotationKeys.length > 0) {
-                        // TODO: this string has newline characters that makes it harder to read
-                        const annotationMapString = annotationKeys
-                            ?.map((a) => `${a.elementID}: ${a.text}`)
-                            .join("\n");
-                        // Provides the annotations for all the element present on screen
-                        // Also provides the annotation of element on which the action needs to be taken
-                        elementAnnotation = await (0, element_annotation_1.getElementAnnotation)({
-                            elementDescription: args.element_description,
-                            annotations: annotationMapString,
-                            annotatedScreenshot: annotatedPageScreenshot,
-                            trace: masterAgentActionSpan,
-                            llm,
-                            preference,
-                        });
-                        output.elementAnnotation = elementAnnotation;
-                        console.log("Output: ", output);
-                        await testGenReporter.sendCurrentView(annotationBuffer);
-                        const triggerHintsFlowSpan = masterAgentActionSpan?.span({
-                            name: "trigger-hints-flow",
-                            input: {
-                                outputFromGetNextAction: output,
-                                generatedAnnotations: annotationKeys,
+                    default: {
+                        const { generatedCodeSteps: codeFromExecuteAction } = await (0, execute_browser_action_1.executeBrowserAction)({
+                            page,
+                            nextAction,
+                            flags: {
+                                useActionSpecificAnnotations,
                             },
-                        });
-                        // Provides the action whether its a click, fill etc.
-                        const result = await (0, with_hints_1.triggerHintsFlow)({
-                            outputFromGetNextAction: output,
-                            generatedAnnotations: annotationKeys,
                             actions,
                             llm,
-                            trace: triggerHintsFlowSpan,
-                        });
-                        shouldTriggerHintsFlow = result.shouldTriggerHintsFlow;
-                        hintsExecutionCompletion = result.hintsExecutionCompletion;
-                        triggerHintsFlowSpan?.end({
-                            output: result,
-                        });
-                    }
-                    if (shouldTriggerHintsFlow && hintsExecutionCompletion) {
-                        const toolCalls = hintsExecutionCompletion?.tool_calls || [];
-                        for (const i in toolCalls) {
-                            const currentToolCall = toolCalls[i];
-                            const code = await actions.executeAction(currentToolCall.function.name, {
-                                ...JSON.parse(currentToolCall.function.arguments),
-                                elementAnnotation,
-                            }, masterAgentActionSpan);
-                            if (code) {
-                                generatedCodeSteps.push(code);
-                            }
-                        }
-                        if (actions.isStuckInLoop()) {
-                            throw new Error("Agent is not able to figure out next action when using hints");
-                        }
-                    }
-                    else {
-                        const browserAction = await (0, browsing_1.executeTaskUsingBrowsingAgent)({
                             trace: masterAgentActionSpan,
-                            action: output.action,
-                            page,
-                            options,
-                            llm,
-                            actions,
                         });
-                        if (browserAction) {
-                            output.action = browserAction.action;
-                            if (browserAction.code) {
-                                generatedCodeSteps.push(browserAction.code);
-                            }
-                        }
+                        generatedCodeSteps.push(...codeFromExecuteAction);
                     }
                 }
                 // resetting error count as there is a successful action
@@ -305,7 +211,7 @@ async function createTestUsingMasterAgent({ task, page, testCase, specPath, opti
                     input: {
                         action: output.action,
                         reason: output.reason,
-                        type: toolCall.actionType,
+                        type: nextAction.actionType,
                     },
                     output: {
                         success: true,
@@ -315,7 +221,7 @@ async function createTestUsingMasterAgent({ task, page, testCase, specPath, opti
                 disableSkills = false;
             }
             catch (e) {
-                if (toolCall.actionType === skill_1.SKILL_USAGE) {
+                if (nextAction.actionType === skill_1.SKILL_USAGE) {
                     logger.log("Disabling skill usage for next retry");
                     disableSkills = true;
                 }
@@ -323,7 +229,7 @@ async function createTestUsingMasterAgent({ task, page, testCase, specPath, opti
                     input: {
                         action: output.action,
                         reason: output.reason,
-                        type: toolCall.actionType,
+                        type: nextAction.actionType,
                     },
                     output: {
                         error: true,
@@ -331,7 +237,7 @@ async function createTestUsingMasterAgent({ task, page, testCase, specPath, opti
                     },
                 });
                 if (!(e instanceof errors_1.HumanApprovalDenied)) {
-                    logger.error("Failed to run master agent's next task", JSON.stringify(toolCall, null, 2), e);
+                    logger.error("Failed to run master agent's next task", JSON.stringify(nextAction, null, 2), e);
                     failedActions.push(output.action);
                     if (failedActions.length >= MAX_ERROR_COUNT) {
                         const error = "Agent is not able to figure out next action, marking task as done";

package/dist/agent/master/scroller.d.ts

@@ -1,5 +1,5 @@
 import { TraceClient } from "@empiricalrun/llm";
-import { Page } from "playwright";
+import type { Page } from "@playwright/test";
 import { CustomLogger } from "../../bin/logger";
 export type FrameReference = {
     scrollPosition: number;

package/dist/agent/master/scroller.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"scroller.d.ts","sourceRoot":"","sources":["../../../src/agent/master/scroller.ts"],"names":[],"mappings":"AAAA,OAAO,EAAuB,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAGrE,OAAO,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AAElC,OAAO,EAAE,YAAY,EAAE,MAAM,kBAAkB,CAAC;AAchD,MAAM,MAAM,cAAc,GAAG;IAC3B,cAAc,EAAE,MAAM,CAAC;IACvB,eAAe,EAAE,MAAM,CAAC;CACzB,CAAC;AA4ZF,wBAAsB,QAAQ,CAAC,EAC7B,kBAAkB,EAClB,IAAI,EACJ,KAAK,EACL,cAAc,EACd,MAAM,GACP,EAAE;IACD,kBAAkB,EAAE,MAAM,CAAC;IAC3B,IAAI,EAAE,IAAI,CAAC;IACX,KAAK,CAAC,EAAE,WAAW,CAAC;IACpB,cAAc,CAAC,EAAE,cAAc,CAAC;IAChC,MAAM,CAAC,EAAE,YAAY,CAAC;CACvB,GAAG,OAAO,CAAC,cAAc,EAAE,CAAC,CA6D5B"}
+{"version":3,"file":"scroller.d.ts","sourceRoot":"","sources":["../../../src/agent/master/scroller.ts"],"names":[],"mappings":"AAAA,OAAO,EAAuB,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAErE,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,kBAAkB,CAAC;AAG7C,OAAO,EAAE,YAAY,EAAE,MAAM,kBAAkB,CAAC;AAchD,MAAM,MAAM,cAAc,GAAG;IAC3B,cAAc,EAAE,MAAM,CAAC;IACvB,eAAe,EAAE,MAAM,CAAC;CACzB,CAAC;AA2ZF,wBAAsB,QAAQ,CAAC,EAC7B,kBAAkB,EAClB,IAAI,EACJ,KAAK,EACL,cAAc,EACd,MAAM,GACP,EAAE;IACD,kBAAkB,EAAE,MAAM,CAAC;IAC3B,IAAI,EAAE,IAAI,CAAC;IACX,KAAK,CAAC,EAAE,WAAW,CAAC;IACpB,cAAc,CAAC,EAAE,cAAc,CAAC;IAChC,MAAM,CAAC,EAAE,YAAY,CAAC;CACvB,GAAG,OAAO,CAAC,cAAc,EAAE,CAAC,CA6D5B"}

package/dist/agent/master/scroller.js

@@ -187,7 +187,6 @@ async function getDivAnnotationToScrollOn({ elementDescription, page, trace, log
         preference: {
             actionType: action_tool_calls_1.ActionType.SCROLL,
         },
-        options: {},
         trace,
     });
     // Remove the used annotations from the list

package/dist/agent/planner/run-time-planner.d.ts

@@ -1,9 +1,10 @@
 import { TraceClient } from "@empiricalrun/llm";
+import type { Page } from "playwright";
 export declare function runtimePlanner({ trace, task, successfulActions, pages, currentPage, }: {
     trace?: TraceClient;
     successfulActions: string[];
     task: string;
-    pages?: Record<string, any>;
+    pages?: Record<string, Page>;
     currentPage?: string;
 }): Promise<{
     pageName: string;

package/dist/agent/planner/run-time-planner.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"run-time-planner.d.ts","sourceRoot":"","sources":["../../../src/agent/planner/run-time-planner.ts"],"names":[],"mappings":"AAAA,OAAO,EAAsB,WAAW,EAAE,MAAM,mBAAmB,CAAC;AAKpE,wBAAsB,cAAc,CAAC,EACnC,KAAK,EACL,IAAI,EACJ,iBAAiB,EACjB,KAAK,EACL,WAAW,GACZ,EAAE;IACD,KAAK,CAAC,EAAE,WAAW,CAAC;IACpB,iBAAiB,EAAE,MAAM,EAAE,CAAC;IAC5B,IAAI,EAAE,MAAM,CAAC;IACb,KAAK,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,GAAG,CAAC,CAAC;IAC5B,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;;;;GAoFA"}
+{"version":3,"file":"run-time-planner.d.ts","sourceRoot":"","sources":["../../../src/agent/planner/run-time-planner.ts"],"names":[],"mappings":"AAAA,OAAO,EAAsB,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACpE,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AAKvC,wBAAsB,cAAc,CAAC,EACnC,KAAK,EACL,IAAI,EACJ,iBAAiB,EACjB,KAAK,EACL,WAAW,GACZ,EAAE;IACD,KAAK,CAAC,EAAE,WAAW,CAAC;IACpB,iBAAiB,EAAE,MAAM,EAAE,CAAC;IAC5B,IAAI,EAAE,MAAM,CAAC;IACb,KAAK,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,IAAI,CAAC,CAAC;IAC7B,WAAW,CAAC,EAAE,MAAM,CAAC;CACtB;;;;GA4FA"}

package/dist/agent/planner/run-time-planner.js

@@ -2,7 +2,7 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.runtimePlanner = void 0;
 const llm_1 = require("@empiricalrun/llm");
-const promptTemplate_0 = "{{#section \"system\"}}\nGiven a successfully executed actions that lists only the actions that were\nsuccessfully executed and a task comprising multiple actions, your goal is to\nanalyse the list and determine if the entire task is completed.\n\nThese actions are executed by AI agents using Playwright on a browser. These agents\nalready have access to browser tabs to execute actions. The successfully executed\nactions on browser post browser has opened, is provided to you as successfully\nexecuted actions.\n\nIf the task is not fully completed, identify which specific actions are missing\nand suggest next steps to complete the task. Assume that the conversation provided\nis entirely truthful and no additional actions were performed beyond those listed.\n\nTo fulfil your goal, follow these steps:\n- Divide the task into individual actions.\n- Compare each task action against the actions listed in the successfully executed actions list.\n- Identify which actions have been executed and which have not.\n- If all actions are executed, respond with the task as done.\n- If any actions are missing, respond with the task as not done, listing all actions\n  and specifying which are complete and which are missing.\n- If provided with list of pages, based on the next pending action and previously executed\n  action, identify the page on which next action needs to be taken    \n{{/section}}\n\n{{#section \"user\"}}\nTask:\n{{task}}\n\n----\n\nFollowing are successfully executed actions:\n{{successfulActions}}\n\n----\n\nCurrent page:\n{{currentPage}}\n\n{{/section}}\n";
+const promptTemplate_0 = "{{#section \"system\"}}\nYou are given a list of successfully executed actions that are done towards completing a task (which\nis also provided to you). Your goal is to analyse the list and determine if the task is completed.\n\nIf the task is not fully completed, identify which specific actions are missing\nand suggest next steps to complete the task. Assume that the conversation provided\nis entirely truthful and no additional actions were performed beyond those listed.\n\nThese actions were executed by AI agents using Playwright on a browser. These agents\nalready have access to browser tabs to execute actions. If there is a pending action,\none of the agents will execute it in the browser. However, they need your help to\nchoose which browser tab (= page) to use for the next action.\n\nTo fulfil your goal, follow these steps:\n- Divide the task into individual actions.\n- Compare each task action against the actions listed in the successfully executed actions list.\n- Identify which actions have been executed and which have not.\n- If all actions are executed, respond with the task as done.\n- If any actions are missing, respond with the task as not done, listing all actions\n  and specifying which are complete and which are missing.\n- If provided with list of pages, based on the next pending action and previously executed\n  action, identify the page on which next action needs to be taken\n{{/section}}\n\n{{#section \"user\"}}\nTask:\n{{task}}\n\n----\n\nSuccessfully executed actions:\n{{successfulActions}}\n\n----\n\nList of pages with their current URLs:\n{{pagesSummary}}\n\n\n{{/section}}\n";
 const utils_1 = require("../utils");
 async function runtimePlanner({ trace, task, successfulActions, pages, currentPage, }) {
     const runTimePlannerSpan = trace?.span({
@@ -14,6 +14,9 @@ async function runtimePlanner({ trace, task, successfulActions, pages, currentPa
         },
     });
     const llm = new llm_1.LLM({ provider: "openai" });
+    const pagesSummary = pages
+        ? Object.entries(pages).map(([pageName, page]) => `${pageName}: Currently on ${page.url()}`)
+        : [];
     const response = await llm.createChatCompletion({
         trace: runTimePlannerSpan,
         traceName: "runtime-planner-llm",
@@ -21,7 +24,7 @@ async function runtimePlanner({ trace, task, successfulActions, pages, currentPa
         messages: (0, llm_1.compilePrompt)(promptTemplate_0, {
             task,
             successfulActions: successfulActions.join("\n"),
-            currentPage,
+            pagesSummary: pagesSummary.join("\n"),
         }),
         tools: [
             {
@@ -44,17 +47,21 @@ async function runtimePlanner({ trace, task, successfulActions, pages, currentPa
                                 type: "string",
                                 description: "reasoning for identification of task status",
                             },
-                            isDone: {
-                                type: "boolean",
-                                description: "whether the task is done",
+                            nextAction: {
+                                type: "string",
+                                description: "next action to be taken",
                             },
                             pageName: {
                                 type: "string",
                                 enum: pages ? Object.keys(pages) : [],
                                 description: "page name for the next action.",
                             },
+                            isDone: {
+                                type: "boolean",
+                                description: "whether the task is done",
+                            },
                         },
-                        required: ["isDone", "reason", "pageName"],
+                        required: ["isDone", "reason", "pageName", "nextAction"],
                     },
                 },
             },

package/dist/evals/master-agent.evals.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"master-agent.evals.d.ts","sourceRoot":"","sources":["../../src/evals/master-agent.evals.ts"],"names":[],"mappings":"AAOA,OAAO,EAAE,UAAU,EAAE,MAAM,QAAQ,CAAC;AAEpC,eAAO,MAAM,4BAA4B,EAAE,UAwC1C,CAAC;AAEF,eAAe,4BAA4B,CAAC"}
+{"version":3,"file":"master-agent.evals.d.ts","sourceRoot":"","sources":["../../src/evals/master-agent.evals.ts"],"names":[],"mappings":"AAMA,OAAO,EAAE,UAAU,EAAE,MAAM,QAAQ,CAAC;AAEpC,eAAO,MAAM,4BAA4B,EAAE,UAoC1C,CAAC;AAEF,eAAe,4BAA4B,CAAC"}

package/dist/evals/master-agent.evals.js

@@ -3,11 +3,10 @@ Object.defineProperty(exports, "__esModule", { value: true });
 exports.masterGetNextActionEvaluator = void 0;
 const actions_1 = require("../actions");
 const skill_1 = require("../actions/skill");
-const action_tool_calls_1 = require("../agent/master/action-tool-calls");
 const next_action_1 = require("../agent/master/next-action");
 const page_1 = require("../page");
 const masterGetNextActionEvaluator = async ({ item, trace, }) => {
-    const { task, executedActions, failedActions, pageUrl, options, pageScreenshot, disableSkills, skills = [], } = item.input;
+    const { task, executedActions, failedActions, options, pageScreenshot, disableSkills, skills = [], } = item.input;
     const page = {};
     skill_1.testCaseSkills.updateSkills(skills);
     const actions = new actions_1.PlaywrightActions(new page_1.TestGenPage(page, "page"));
@@ -15,12 +14,10 @@ const masterGetNextActionEvaluator = async ({ item, trace, }) => {
         task,
         executedActions,
         failedActions,
-        pageUrl,
         trace,
         options,
         pageScreenshot,
         actions,
-        actionTypes: action_tool_calls_1.ActionType,
         disableSkills,
         page,
     });

package/dist/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AAOlC,OAAO,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAYpC,wBAAsB,UAAU,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,IAAI,EAAE,KAAK,CAAC,EAAE,SAAS,iBAsC3E"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AACA,OAAO,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AAOlC,OAAO,EAAE,SAAS,EAAE,MAAM,SAAS,CAAC;AAYpC,wBAAsB,UAAU,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,IAAI,EAAE,KAAK,CAAC,EAAE,SAAS,iBAiC3E"}

package/dist/index.js

@@ -40,11 +40,6 @@ async function createTest(task, page, scope) {
         page,
         task,
         options: {
-            htmlSanitize: {
-                disallowedStrings: [
-                    "v-data-table__td v-data-table-column--align-start",
-                ],
-            },
             ...testGenConfig.options,
         },
         scopeVars: scope,

package/dist/page/index.d.ts

@@ -1,4 +1,4 @@
-import { Page } from "playwright";
+import type { Page } from "playwright";
 export declare class TestGenPage {
     pwPageInstance: Page;
     name: string;

package/dist/page/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/page/index.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AAElC,qBAAa,WAAW;IAEb,cAAc,EAAE,IAAI;IACpB,IAAI,EAAE,MAAM;gBADZ,cAAc,EAAE,IAAI,EACpB,IAAI,EAAE,MAAM;IAGrB,UAAU,CAAC,EAAE,IAAI,EAAE,IAAI,EAAE,EAAE;QAAE,IAAI,EAAE,IAAI,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE;CAIxD"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../../src/page/index.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,YAAY,CAAC;AAEvC,qBAAa,WAAW;IAEb,cAAc,EAAE,IAAI;IACpB,IAAI,EAAE,MAAM;gBADZ,cAAc,EAAE,IAAI,EACpB,IAAI,EAAE,MAAM;IAGrB,UAAU,CAAC,EAAE,IAAI,EAAE,IAAI,EAAE,EAAE;QAAE,IAAI,EAAE,IAAI,CAAC;QAAC,IAAI,EAAE,MAAM,CAAA;KAAE;CAIxD"}

package/dist/utils/env.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"env.d.ts","sourceRoot":"","sources":["../../src/utils/env.ts"],"names":[],"mappings":"AAAA,wBAAgB,gBAAgB,YAE/B"}
+{"version":3,"file":"env.d.ts","sourceRoot":"","sources":["../../src/utils/env.ts"],"names":[],"mappings":"AAAA,wBAAgB,gBAAgB,YAM/B"}

package/dist/utils/env.js

@@ -2,6 +2,8 @@
 Object.defineProperty(exports, "__esModule", { value: true });
 exports.isRunningOnCloud = void 0;
 function isRunningOnCloud() {
-    return process.env.CI === "true";
+    return (process.env.CI === "true" ||
+        // eslint-disable-next-line turbo/no-undeclared-env-vars
+        !!process.env.RUNNING_BROWSER_TESTS_FOR_TEST_GEN);
 }
 exports.isRunningOnCloud = isRunningOnCloud;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@empiricalrun/test-gen",
-  "version": "0.46.7",
+  "version": "0.46.9",
   "publishConfig": {
     "registry": "https://registry.npmjs.org/",
     "access": "public"

package/playwright.config.ts

@@ -8,7 +8,7 @@ export default defineConfig({
   testDir: "./dist/.",
   testMatch: "**/*.spec.js",
   // Setting large timeout as master agent takes time to run.
-  timeout: 240_000,
+  timeout: 300_000,
   fullyParallel: true,
   workers: "50%",
   use: {

package/dist/agent/browsing/o1-completion.d.ts

@@ -1,8 +0,0 @@
-import { TraceClient } from "@empiricalrun/llm";
-import { ChatCompletionMessage, ChatCompletionMessageParam, ChatCompletionTool } from "openai/resources/index.mjs";
-export declare function getO1Completion({ messages, tools, trace, }: {
-    messages: ChatCompletionMessageParam[];
-    tools: ChatCompletionTool[];
-    trace?: TraceClient;
-}): Promise<ChatCompletionMessage | undefined>;
-//# sourceMappingURL=o1-completion.d.ts.map

package/dist/agent/browsing/o1-completion.d.ts.map

@@ -1 +0,0 @@
-{"version":3,"file":"o1-completion.d.ts","sourceRoot":"","sources":["../../../src/agent/browsing/o1-completion.ts"],"names":[],"mappings":"AAAA,OAAO,EAAO,WAAW,EAAE,MAAM,mBAAmB,CAAC;AACrD,OAAO,EACL,qBAAqB,EACrB,0BAA0B,EAC1B,kBAAkB,EACnB,MAAM,4BAA4B,CAAC;AAMpC,wBAAsB,eAAe,CAAC,EACpC,QAAQ,EACR,KAAK,EACL,KAAK,GACN,EAAE;IACD,QAAQ,EAAE,0BAA0B,EAAE,CAAC;IACvC,KAAK,EAAE,kBAAkB,EAAE,CAAC;IAC5B,KAAK,CAAC,EAAE,WAAW,CAAC;CACrB,GAAG,OAAO,CAAC,qBAAqB,GAAG,SAAS,CAAC,CA8D7C"}

package/dist/agent/browsing/o1-completion.js

@@ -1,72 +0,0 @@
-"use strict";
-var __importDefault = (this && this.__importDefault) || function (mod) {
-    return (mod && mod.__esModule) ? mod : { "default": mod };
-};
-Object.defineProperty(exports, "__esModule", { value: true });
-exports.getO1Completion = void 0;
-const llm_1 = require("@empiricalrun/llm");
-const remove_markdown_1 = __importDefault(require("remove-markdown"));
-const constants_1 = require("../../constants");
-const utils_1 = require("../utils");
-async function getO1Completion({ messages, tools, trace, }) {
-    let completion;
-    try {
-        const o1Span = trace?.span({ name: "o1-response-span" });
-        const llm = new llm_1.LLM({
-            trace: o1Span,
-            provider: "openai",
-            defaultModel: "o1-mini",
-            providerApiKey: constants_1.MODEL_API_KEYS["openai"],
-        });
-        const [userInstruction] = messages.filter((s) => s.role === "user");
-        const [systemInstruction] = messages.filter((s) => s.role === "system");
-        userInstruction.content = `${systemInstruction?.content} 
-  
-    ${userInstruction?.content}
-
-    You need to respond with one of the following tool call with provided schema:
-    ${tools.map((tool) => JSON.stringify(tool, null, 2)).join("\n --- \n")}
- 
-    ------
-
-    Before responding, ensure the following:
-    - Do not respond with markdown, respond only with the JSON object.
-    - Do not respond with any backticks.
-    - The reason for action should also include what was been executed in the action.
-  `;
-        o1Span?.update({ input: [userInstruction] });
-        const response = (await llm.createChatCompletion({
-            messages: [userInstruction],
-            modelParameters: {
-                ...constants_1.DEFAULT_O1_MODEL_PARAMETERS,
-            },
-        }));
-        o1Span?.end({ output: response });
-        const toolResponseStr = (0, remove_markdown_1.default)(response.content);
-        const toolRespJSON = (0, utils_1.parseJson)(toolResponseStr);
-        const parameters = toolRespJSON.function.parameters || toolRespJSON.function.arguments;
-        if (!parameters) {
-            throw new Error("No parameters found in tool response");
-        }
-        const tool = {
-            type: "function",
-            function: {
-                name: toolRespJSON.function.name,
-                arguments: typeof parameters === "string"
-                    ? parameters
-                    : JSON.stringify(parameters, null, 2),
-            },
-        };
-        completion = {
-            role: "assistant",
-            content: response.content,
-            tool_calls: [tool],
-        };
-    }
-    catch (e) {
-        console.error("O1 response error", e);
-        return undefined;
-    }
-    return completion;
-}
-exports.getO1Completion = getO1Completion;